Letter:Statistics ensure good results

Sir: I agree with Dr Bird that statistics is the servant of science, not its ruler (Letters, 1 June). However good statistical practice provides a formalised way of ensuring that good science is done.

The purpose of randomisation blinding and use of placebos in clinical trials is to avoid the conscious and unconscious biases which can occur when patients or clinical investigators use other means of selecting treatment. In addition randomisation protects the public, in permitting the rigorous estimation of probabilities of obtaining a false positive or false negative result from a particular trial design.

With regard to Dr Bird's alternatives of comparing non-randomised trial results with historical data or "a scientific estimate of the placebo effect", if such estimates are based on data they are inherently statistical - if not based on data are they scientific?

DAVID MORGAN

Wokingham, Berkshire